HDAC3 Deacetylates the DNA Mismatch Repair Factor Mutsβ to Stimulate Triplet Repeat Expansions

HDAC3 deacetylates the DNA mismatch repair factor MutSβ to stimulate triplet repeat expansions

Gregory M. Williamsa, Vasileios Paschalisb, Janice Ortegac, Frederick W. Muskettb, James T. Hodgkinsond, Guo-Min Lic, John W. R. Schwabeb, and Robert S. Lahuea,e,1

aCentre for Chromosome Biology, National University of Ireland Galway, H9W2TY Galway, Ireland; bLeicester Institute of Chemical and Molecular Biology, Department of Molecular and Cell Biology, University of Leicester, LE1 7RH Leicester, United Kingdom; cDepartment of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX 75390; dLeicester Institute of Chemical and Molecular Biology, School of Chemistry, University of Leicester, LE1 7RH Leicester, United Kingdom; and eGalway Neuroscience Centre, National University of Ireland Galway, H9W2TY Galway, Ireland

Edited by Philip C. Hanawalt, Stanford University, Stanford, CA, and approved August 17, 2020 (received for review June 30, 2020) Trinucleotide repeat (TNR) expansions cause nearly 20 severe genome (25, 26). This fact suggests a highly localized mutagenic human neurological diseases which are currently untreatable. event that only affects the triplet repeat itself (7–10). These unusual For some of these diseases, ongoing somatic expansions accelerate features of triplet repeat expansions point to unique mechanisms of disease progression and may influence age of onset. This new mutagenesis. knowledge emphasizes the importance of understanding the The DNA mismatch repair complex MutSβ (Msh2-Msh3 het- protein factors that drive expansions. Recent genetic evidence erodimer) has been identified as a main driver of both inherited indicates that the mismatch repair factor MutSβ (Msh2-Msh3 com- and somatic triplet repeat expansions. Gene knockouts in HD, plex) and the histone deacetylase HDAC3 function in the same DM1, and FXS mice clearly demonstrate that both MSH2 and pathway to drive triplet repeat expansions. Here we tested the MSH3 are required for virtually all expansions (20, 21, 27–34). In hypothesis that HDAC3 deacetylates MutSβ and thereby activates addition, naturally occurring single nucleotide polymorphisms it to drive expansions. The HDAC3-selective inhibitor RGFP966 was (SNPs) in mouse MSH3 also have clear effects on somatic ex- used to examine its biological and biochemical consequences in pansions, with lower expansion frequencies correlating with lower human tissue culture cells. HDAC3 inhibition efficiently suppresses Msh3 protein abundance (35). In contrast, the related mismatch repeat expansion without impeding canonical mismatch repair ac- repair complex MutSα (Msh2-Msh6 heterodimer) appears to play tivity. Five key lysine residues in Msh3 are direct targets of HDAC3 almost no role in expansions (28, 30–34, 36). Experiments in yeast BIOCHEMISTRY deacetylation. In cells expressing Msh3 in which these lysine resi- (37, 38) and human cell culture (39–42) support the primary role dues are mutated to arginine, the inhibitory effect of RGFP966 on of MutSβ in driving expansions. Together, the results from model expansions is largely bypassed, consistent with the direct deace- systems pinpoint MutSβ as a major driving force for expansions. β tylation hypothesis. RGFP966 treatment does not alter MutS sub- The focus on MutSβ was recently underscored by genomewide unit abundance or complex formation but does partially control its association studies (GWAS) of patients with HD or other subcellular localization. Deacetylation sites in Msh3 overlap a nu- CAG•CTG repeat disorders. Disease age of onset and rate of β clear localization signal, and we show that localization of MutS is progression are significantly influenced by polymorphisms in MSH3 partially dependent on HDAC3 activity. Together, these results in- – β (43 46) and other DNA repair genes (47, 48). A polymorphism in dicate that MutS is a key target of HDAC3 deacetylation and MSH3 is also associated with levels of somatic instability of the provide insights into an innovative regulatory mechanism for triplet repeat expansions. The results suggest expansion activity may be druggable and support HDAC3-selective inhibition as an attrac- Significance tive therapy in some triplet repeat expansion diseases. Triplet repeat expansions cause nearly 20 inherited neurological histone deacetylase 3 | mismatch repair | triplet repeat expansion diseases. The molecular mechanism of expansions is of interest at two levels. First is the basic science of how triplet repeats expand, and the second is the therapeutic opportunity to treat riplet repeats are tandem runs of three nucleotides that occur expansion diseases. This study presents evidence that addresses naturally in the genome. Most triplet repeats are genetically T both areas of interest. We show that the histone deacetylase stable and undergo relatively few expansions (gain of repeats) or HDAC3 acts directly on the DNA mismatch repair protein MutSβ contractions (loss of repeats). Some triplet repeats are known to to stimulate CAG•CTG repeat expansions. This basic science in- expand and cause nearly 20 human genetic diseases, including ’ sight into the mechanism also provides an explanation for how Huntington s disease (HD) (1), myotonic dystrophy type 1 (DM1) HDAC3 inhibition in Huntington’s disease mice leads to sup- – (2 4), and fragile X syndrome (FXS) (5, 6). Expansions can occur pression of triplet repeat expansions and could therefore be both between generations (inherited expansions) and, for some considered as a therapy for Huntington’s disease and potentially diseases, within tissues of an individual (somatic expansions) other CAG•CTG repeat expansion diseases. (reviewed in refs. 7–10). The medical importance of triplet repeat expansions has generated significant interest in understanding the Author contributions: G.M.W., V.P., J.O., G.-M.L., J.W.R.S., and R.S.L. designed research; mechanisms of expansion and in the potential for therapies that G.M.W., V.P., J.O., and R.S.L. performed research; G.M.W., V.P., F.W.M., and J.T.H. con- tributed new reagents/analytic tools; G.M.W., V.P., J.O., F.W.M., J.T.H., G.-M.L., J.W.R.S., might modulate somatic expansions to alter disease pathology and R.S.L. analyzed data; and G.M.W., V.P., J.O., F.W.M., J.T.H., G.-M.L., J.W.R.S., and R.S.L. (9, 10). wrote the paper. Analysis of expansions in humans and mice revealed several The authors declare no competing interest. unique features. First, the frequency of expansions can be very high, This article is a PNAS Direct Submission. sometimes approaching 100% (11). Second, expansions can occur This open access article is distributed under Creative Commons Attribution-NonCommercial- both in proliferating tissues (12–16) and in nonproliferating tissues NoDerivatives License 4.0 (CC BY-NC-ND). (17–22). This suggests that at least some expansions arise inde- 1To whom correspondence may be addressed. Email: [email protected]. pendently of cellular DNA replication (23, 24). Third, disease- This article contains supporting information online at https://www.pnas.org/lookup/suppl/ causing expansions appear to be limited to the disease gene locus, doi:10.1073/pnas.2013223117/-/DCSupplemental. with no evidence of significant genetic instability elsewhere in the

www.pnas.org/cgi/doi/10.1073/pnas.2013223117 PNAS Latest Articles | 1of9 Downloaded by guest on September 27, 2021 expanded CTG repeat in DM1 patients (49). The human GWAS (Invitrogen-Thermo Fisher Scientific) (SI Appendix,Fig.S1A). The transfected findings have triggered substantial interest in therapies that target cells were treated with either the HDAC3-selective inhibitor RGFP966 from the ability of MutSβ to promote somatic expansions as a means of BioMarin Pharmaceutics, Inc (56, 57) or DMSO (dimethyl sulfoxide; Sigma- μ μ slowing or preventing pathology. Aldrich) at a final concentration of 1 Mor10 Mandincubatedfor48hto Proteins that potentially regulate MutSβ also drive triplet re- allow for potential expansion of the repeat tract. Cells were then harvested and the shuttle vector plasmid recovered and digested with the restriction peat expansions. The role of histone deacetylases (HDACs) in enzyme DpnI (New England Biolabs). The digested shuttle vector DNA is then triplet repeat expansion was first identified in yeast and subse- transformed into Saccharomyces cerevisiae where a genetic selection (His+ quently in human cells (50). siRNA knockdown or chemical in- CanR) identifies those plasmid molecules with an apparent expansion. Colonies hibition of the human class I enzyme HDAC3 resulted in loss of from the selective plates are then investigated by colony PCR to visualize, most expansions. Knockdown of the class IIa HDAC5 or double validate, and size the expansions (55). knockdowns of HDAC3 and HDAC5 also suppressed expansions (39). The genetic epistasis between HDAC3 and HDAC5 is Subcellular Fractionation and Sucrose Cushion Experiments. Subcellular frac- consistent with the known components of the transcriptional tionation steps were adapted from Dignam et al. (58). SVG-A cells were seeded corepressor SMRT/NCoR (51), which contains both HDAC3 in T175 flasks (Sarstedt) and incubated for 24 h before treatment with and HDAC5 (52). Since HDAC3 has the dominant deacetylase RGFP966. After 24 h, treated cells were trypsinized, collected, and washed × activity, it is likely the complex acts by deacetylating one or more twice with ice cold 1 PBS. All subsequent steps were carried out at 4 °C. Cells × × protein targets that lead to enhanced triplet repeat expansions. were then washed with 5 cell pellet volume of 1 buffer A (10 mM Hepes pH 7.9, 1.5 mM MgCl , 10 mM KCl, 0.5 mM dithiothreitol 1× concentration of The counteracting histone acetyl transferase (HAT) appears to 2 mammalian protease inhibitors [Fisher Scientific 12841640]) and centrifuged at be CBP/p300, based on the finding that siRNA knockdown of 450 × g for 5 min. Supernatant was discarded and the cell pellets were these HATs stimulated expansions (50). resuspended in 2× cell pellet volume of buffer A and put on ice for 20 min to β A functional connection between MutS and HDAC3/SMRT swell. After incubation, an equal volume of buffer A + 1.0% Nonidet P-40 is highlighted by genetic data indicating that these factors work (Sigma-Aldrich) was added to the cells and incubated on ice for 5 min. Nuclei in the same pathway to promote CAG•CTG repeat expansion were collected by centrifugation at 450 × g for 10 min, and the supernatant (39). This idea is further supported by studies in HD mice, where (cytoplasmic fraction) was saved for protein quantification and immunoblot treatment with the HDAC3-selective inhibitor RGFP966 signif- analysis. Recovered nuclei were washed several times with buffer A and icantly reduced striatal CAG repeat expansions in the HTT gene centrifuged at 450 × g for 5 min. Washed pellets were then resuspended in (53). The evidence from mammalian systems suggests that 1 mL of buffer A and layered over 10 mL of 30% sucrose cushion (10 mM Pipes × HDAC3/SMRT deacetylation might activate MutSβ directly, pH 7.0, 10 mM KCl, 1.5 mM MgCl2, 30% wt/vol sucrose, 1 concentration of × resulting in expansions. However, there have been no data to mammalian protease inhibitors) before being centrifuged at 800 g for 10 min. The supernatant was discarded, and the previous step repeated. The evaluate this suggestion. In this study, we investigated the hy- β remaining nuclei were resuspended in 1 mL of buffer A and spun down at pothesis that the expansion activity of MutS is directly con- 450 × g for 5 min, washed again in 1× PBS, and centrifuged at 450 × g for trolled through its acetylation state, and that deacetylation by 5 min. The supernatant was discarded and the remaining nuclei (nuclear HDAC3/SMRT activates MutSβ to drive expansions. The hy- fraction) were lysed with 50 to 200 μL of radioimmunoprecipitation assay lysis pothesis was evaluated by genetic, biochemical, and pharmaco- buffer (see Immunoblot Analysis above) and subjected to protein quantifica- logical approaches in cultured human cells and with purified tion and immunoblot analysis. components. Expansions in this system require MutSβ (39, 42); they show specificity for expandable repeat sequences over Creation of Msh3-5KR Cell Line. The Msh3-5KR variant was generated by site- nonexpandable sequence (CAG repeats versus TAG repeats) directed mutagenesis (Q5 site-directed mutagenesis kit, New England Biolabs) −/− + (54); and there is a MSH3 derivative in which variant Msh3 using the Msh3 expression plasmid pGWB20 (Msh3-cDNA in pcDNA3.1 Neo ) + proteins can be expressed without interference from wild-type to create pGWB21 (Msh3-5KR-cDNA in pcDNA3.1 Neo ). The mutated plasmid α Msh3 (42). was transformed into NEBStable DH5 Escherichia coli (New England Biolabs) and positive pGWB21 candidates were selected for by ampicillin resistance. Materials and Methods Mutations were confirmed by Sanger sequencing (Source Bioscience). SVG-A Msh3 −/− cells were cotransfected with pGWB21 Msh3-5KR expression plasmid Immunoblot Analysis. Human SVG-A cells were harvested by washing with and a pMSCV hygro resistance plasmid (Clontech Laboratories) using Lip- ice-cold PBS (phosphate-buffered saline) and resuspended in lysis buffer ofectamine 2000 (Thermo Fisher). After a 48-h incubation, cells were resus- (0.1% wt/vol SDS [sodium dodecyl sulfate], 0.1% vol/vol Triton X-100, 0.1% pended in hygromycin-supplemented media (400 ng/mL) and diluted to 200 wt/vol sodium deoxylcholate, 5 mM ethylenediaminetetraacetic acid, cells/10 cm2 dish to allow formation of colonies. The 1× expression of Msh3- 100 mM phenylmethylsulfonyl fluoride) plus mammalian protease inhibitors 5KR was determined by immunoblot analysis, compared to wild-type Msh3 (Fisher Scientific 12841640) at 1× concentration. Following a 30-min incu- expression using two different antibodies, BD Biosciences 611390 and bation on ice, cells were sonicated and centrifuged at 16,200 × g at 4 °C for Abcam 154251. 40 min. Supernatant protein concentrations were quantified using the de- tergent compatible protein assay (Bio-Rad Laboratories). Samples containing 91-130 50 to 100 μg of total protein were denatured, then resolved on either 10% Creation of Msh3-EGFP and SV40 NLS3-EGFP Fusion Plasmids. Msh3 -EGFP 91-130 SDS/polyacrylamide gel electrophoresis gels (Msh2, Msh3, actin, α-tubulin, and Msh3-5KR -EGFP fusion proteins were created from plasmids pGWB20 + HDAC3) or precast 4 to 20% gradient gels (Bio-Rad 4565093) (histone H4, (Msh3-cDNA in pcDNA3.1 Neo )andpGWB21 (Msh3-5KR-cDNA in pcDNA3.1 + acetylated histone H4). Following transfer to nitrocellulose membrane and Neo ). Briefly, PCR was used to amplify the MSH3 or MSH3-5KR coding region blocking, membranes were incubated at 4 °C overnight with 1:1,000 dilu- from codons 91 to 130 using primers Msh3 91 to 130F and Msh3 91 to 130R (SI tions of primary antibodies: Msh3, BD Biosciences 611390 or Abcam 154251; Appendix,Fig.S1B). Following cleavage with XhoIandAgeI, the PCR products Msh2, Calbiochem NA26; actin, Sigma-Aldrich A2066; histone H4, Abcam were cloned into the corresponding sites of vector pEGFP-N1 (originally from AB31830; acetylated histone H4, Millipore, 06-866; and α-tubulin, Abcam Clontech; a generous gift of Kevin Sullivan, National University of Ireland, AB6161-100. After washing, membranes were incubated for 1 h with Galway, Ireland). The SV40 nuclear localization signal (NLS) was used as a 1:10,000 dilutions of secondary antibodies (Li-Cor Biosciences) IRDye 800CW positive control for nuclear localization. We created an SV40-NLS3x-EGFP fu- goat α-mouse IgG P/N 925-32210, or IRDye 800CW goat α-rabbit IgG P/N 925- sion, with three tandem copies of the NLS sequence PKKKRKV, by enzymatic 3211 at room temperature. Visualization and quantification was performed phosphorylation of oligonucleotides SV40 NLS3 F and SV40 NLS3 R, followed using the Odyssey Infrared Imaging system and software (Li-Cor Biosciences). by heating and slow cooling to allow annealing. The duplex product was li- gated into the XhoIandAgeI sites of vector pEGFP-N1. To create full-length Shuttle Vector Assay and Measurement of Trinucleotide Repeat (TNR) Expansion Msh3-EGFP and Msh3-5KR-EGFP constructs, PCR was performed on the full Frequencies. The SVG-A shuttle vector assay has been previously described (55). coding region of MSH3 or MSH3-5KR coding region using primers Msh3 Gib-

Briefly, a shuttle vector harboring a (CAG)22 repeat upstream of a CAN1 re- son F and Msh3 Gibson R, followed by Gibson assembly in pEGFP-N1. All porter gene is transfected into human SVG-A cells using Lipofectamine 2000 plasmid clones were confirmed by DNA sequencing.

2of9 | www.pnas.org/cgi/doi/10.1073/pnas.2013223117 Williams et al. Downloaded by guest on September 27, 2021 Generation of GFP-Msh3 Stable Cell Lines. SVG-A cells were transfected with S1C and, where statistical significance was reached, in the relevant figure Lipofectamine 2000 (Thermo Fisher) using 500 ng of either full-length GFP- legend. Statistical tests used for each P value are stated for each figure. Msh3 or full-length GFP-Msh3-5KR plasmids and incubated for 48 h. Trans- fected cells were then subjected to limited dilution (one cell per five wells in Results 96-well dishes) to obtain a monoclonal cell population of GFP-positive can- HDAC3 Targets Specific Lysine Residues in Msh3 to Stimulate Triplet didates. The 96-well plates were incubated for 14 to 21 d to allow single cells to β grow into colonies. GFP-positive colonies were then bulked up to 25 mm2 flasks Repeat Expansions. If acetylation and deacetylation of MutS where stable integrations could be confirmed by PCR (SI Appendix,Fig.S1B)GFP- control its ability to catalyze triplet repeat expansions, then there Msh3 protein expression was measured in individual clones by immunoblot should be specific lysine residues in MutSβ which are the targets analysis using α-Msh3 (BD Biosciences 611390) and α-GFP antibodies.

Immunofluorescence Microscopy. The 1.2 × 105 SVG-A cells were seeded in six- well tissue culture plates (Sarstedt) containing sterile glass coverslips (Fisher) and 2 mL of complete Dulbecco’s modified Eagle media and incubated over-

night at 37 °C and 5% CO2.After∼16 h, cells were treated with either DMSO or RGFP966 at a final concentration of 10 μM for 24 h. Cells were briefly washed with warm 1× PBS, then fixed with warm 4% vol/vol paraformalde- hyde (Sigma) for 10 min at 37 °C. After fixation, coverslips were washed again with 1× PBS and air dried. A total of 10 μLofSlowFadeGoldplusDAPI (Thermo Fisher) was used to mount each coverslip. Z-stack images were acquired using an Olympus Delta Vision microscope at either 100× or 60× magnification with a 0.5-μm limit per image. Images were processed using Fiji ImageJ software.

Mismatch Repair Assay. HeLa cells were grown to 90% confluence in RPMI (Roswell Park Memorial Institute) 1640 media supplemented with 5% fetal bovine serum. RGFP966 (5 μM) was added to cells 12 h before harvesting. Nuclear extracts were prepare as previously described (59). DNA substrates used for the mismatch repair (MMR) assay are circular heteroduplex containing a single G-T or a 2-nucleotide insertion/deletion (2-ID) mismatch and a strand break 5′ to the DNA lesion (60, 61). In vitro MMR assays were

performed as described (62). Briefly, a 20-μL reaction containing 30 fmol BIOCHEMISTRY DNA substrate and 75 μg nuclear extract was incubated for 15 min at 37 °C. DNA was recovered and digested with NsiI/ClaI and XmcI/ClaI for the G-T and 2-ID substrates, respectively. Reaction products were visualized by ethidium bromide staining. MMR products were quantified using ImageJ.

DNA Hairpin Repair Assay. HeLa cells were again grown to 90% confluence in RPMI 1640 media supplemented with 5% fetal bovine serum, and RGFP966 (5 μM) was added to cells 12 h before harvesting. Nuclear extracts were prepare as previously described (59). DNA substrates containing a loop or hairpin of 5 or 15 extra CAG repeats were prepared as described (63). Hairpin repair assay was performed in a 40-μL reaction containing 42 fmol of DNA substrate and 100 μg of nuclear extract. Reactions were incubated at 37 °C for 30 min. DNA was recovered and digested using BglI and PstI before loading to a 6% denaturing polyacrylamide gel to resolve the products. Repair products were visualized by Southern hybridization using a 32P-labeled probe specifically annealing to the nick (repair) strand near the PstI site (see Fig. 3C). Repair activity was quantified using ImageJ. Graph and statistical analysis was obtained by t test analysis using GraphPad Prism.

NMR Experiments. Acetylated peptides (SI Appendix,Fig.S2)weredissolvedin thesamebufferasthegelfiltrationbufferusedtopurifytheSMRT:HDAC3 complex (10 mM Tris·HCl pH 7.5, 50 mM NaCl) at a concentration of 1.0 mg/mL. NMR samples were prepared containing 200 μM of peptide, 122 nM of

SMRT(389-480)/HDAC3 complex, and 8 μMofIP6 in 5% vol/vol D2O(600μL final volume in a 5-mm Wilmad NMR tube). In vitro deacetylation of the peptide substrates was monitored in real time using successive 1Hone- dimensional (1D) spectra. We recorded a reference 1H 1D spectrum in the absence of the complex, by performing four dummy scans and 16 actual scans, leading to a total acquisition time of 61.7 s. The SMRT(389-480)/HDAC3 complex was added to initiate the reaction. A series of 1H 1D spectra with the same parameters with the reference spectrum were recorded for 60 min. The time between the addition of the complex and the acquisition of the first 1H1D spectrum was monitored (68 s to 81 s, depending on the peptide) was noted down for each experiment. The 1D NMR spectra were acquired using a Bruker Avance 600 MHz. The NMR spectra were processed using “TopSpin” (Bruker). Fig. 1. Putative HDAC3 sites in MutSβ affect triplet repeat expansions. (A) The integrals of the peaks of interest were determined using the “Kinetics Previous proteomic studies (https://www.phosphosite.org//homeAction. Method” in the program “Dynamics Center” (Bruker). As the integral of a action) identified five acetylated lysine residues in Msh3. The Msh3-5KR peak correlates with the concentration of the protons generating the peak, mutant changes these lysine residues to arginines, mimicking a perma- we were able to quantify the percentage of product over the initial substrate nently deacetylated Msh3 protein. (B) Representative Western blot after using the equation in SI Appendix,Fig.S3A. HDAC3 inhibition by RGFP966 treatment. (C) Quantification of Western blot signals as a function of RGFP966 dose. **P < 0.001, n = 4. (D) TNR expan- Statistical Analysis. All P values and n values for repeat instability, Western sions as a function of RGF966 dose in cells expressing Msh3 or Msh3-5KR. blot, and immunoprecipitation experiments are specified in SI Appendix, Fig. *P < 0.05, n = 4. See SI Appendix, Fig. S1C for statistical details.

Williams et al. PNAS Latest Articles | 3of9 Downloaded by guest on September 27, 2021 for this regulation. Information in a proteomics database (https:// relative rates are in general accordance with published data for www.phosphosite.org//homeAction.action) listed acetylated ly- SMRT/HDAC3 preferences for the two amino acids that follow sine residues, originally identified by proteomics studies (sum- the acetylated lysine (66). Thus, SMRT(389-480)/HDAC3 is able marized in ref. 64). We focused on the Msh3 subunit since it is to bind and deacetylate lysine residues of Msh3 peptides. These specific to MutSβ. Five acetylation sites are clustered in the N results support the model where SMRT/HDAC3 deacetylates key terminus of Msh3 (residues K98, K99, K103, K122, and K123) lysine residues in Msh3 to regulate its expansion activity. (Fig. 1A). All five lysine residues were mutated to arginine to create the variant protein Msh3-5KR (Fig. 1A) and thereby block Mismatch Repair Activity Is Unaffected by HDAC3 Inhibition. If HDAC3 acetylation at the five residues. Our hypothesis predicts that deacetylates MutSβ to activate it for triplet repeat expansions, is Msh3-5KR protein will mimic the deacetylated form of MutSβ canonical mismatch repair activity also affected? This question was and, therefore, cells expressing Msh3-5KR will bypass the re- addressed using biochemical assays for mismatch repair. HeLa cells quirement for deacetylation by HDAC3. The plasmid with were treated with DMSO as control or 10 μMRGFP966,then −/− MSH3-5KR was stably integrated into the genome of MSH3 nuclear extracts were prepared and analyzed. Immunoblots indi- cells (42) and integrants were identified that expressed Msh3- cated efficacy of HDAC3 inhibition as histone H3 acetylation was 5KR at near wild-type levels (1.1× normal Msh3 level; SI increased at lysine residues 9, 14, and 18 (Fig. 3A). The abundance Appendix, Fig. S4 A–D). of mismatch repair proteins Msh2, Msh3, Msh6, Mlh1, and Pms2 The HDAC3-selective inhibitor RGFP966 was used to block appeared unaffected by RGFP966 treatment. When the extracts the putative deacetylation of MutSβ. Control experiments showed were assayed for mismatch repair activity, the RGFP966-treated that the inhibitor was efficacious, with a dose-dependent increase sample showed as much or more mismatch repair on 5′ nicked in acetylated histone H4, but with no significant effect on the substrates containing a G-T mismatch or a 2-nt insertion/deletion abundance of either HDAC3 itself or total H4 (Fig. 1 B and C and loop (Fig. 3B). Similarly, alternative substrates with loops of 5 or 15 SI Appendix C , Fig. S1 ). In triplet repeat expansion assays with cells CAG repeats (Fig. 3C) showed as much or more repair to the loop- expressing wild-type Msh3, RGFP966 showed a dose-dependent retention product in extracts from RGFP966-treated cells (Fig. 3 D D SI Appendix C inhibition of expansions (Fig. 1 and ,Fig.S1 ). and E and SI Appendix,Fig.S1C). By these criteria, HDAC3 in- Reductions in expansion frequency by approximately 50% and hibition does not interfere with canonical mismatch repair or CAG μ 85% of control values were observed at RGFP966 doses of 1 M loop repair. and 10 μM, respectively. The latter expansion value is similar in magnitude to that reported previously when HDAC3 expression MutSβ Contains a Potential NLS That Is Sensitive to HDAC3 Activity. was ablated with siRNA (39), supporting the premise that Several initial possibilities as to how HDAC3 might regulate RGFP966 reduces expansions through inhibition of HDAC3. MutSβ were tested and rejected. One idea is that HDAC3 regu- Control experiments indicated that RGFP966 treatment had little lates transcription of MSH2 and/or MSH3, so inhibition of SI Appendix effect on cell viability or proliferation ( , Fig. S5), HDAC3 would alter gene expression and protein abundance. eliminating the possibility that the inhibitor reduces cell numbers Steady-state levels of Msh2, wild-type Msh3, and Msh3-5KR were and thereby reduces expansion frequency. examined by immunoblot of extracts prepared from SVG-A cells When expansions were tested in cells expressing Msh3-5KR and SI D treated with DMSO alone or increasing doses of RGFP966 ( treated with RGFP966, a different pattern emerged (Fig. 1 ). In Appendix,Fig.S4A–D). The results show no significant difference the absence of inhibitor, the expansion activity was increased in abundance of Msh2, wild-type Msh3, or Msh3-5KR. This compared to cells expressing wild-type Msh3, although the in- finding is in accord with earlier studies where Msh2 and Msh3 crease was not statistically significant (P = 0.2215; SI Appendix, C protein levels were unaffected by siRNA knockdown of HDAC3 Fig. S1 ). The increase is consistent with the idea that Msh3-5KR (39). A second possibility is that HDAC3 regulates MutSβ het- alters the putative acetylation state of MutSβ toward the deace- erodimer formation or stability. Coimmunoprecipitation (co-IP) tylated form and therefore causes more expansions. Further dif- experiments showed there was no significant difference in levels of ferences were observed when RGFP966 was added. While about MutSβ complex in extracts of wild-type cells treated with DMSO one-half of the expansions were still sensitive to the inhibitor, the only or 10 μM RGFP966 (SI Appendix, Fig. S4 E–H). Co-IP ex- other ∼50% of expansions were resistant to HDAC3 inhibition. periments from cells expressing Msh3-5KR showed a slight de- This effect is most apparent for RGFP966 treatment at 10 μM crease in MutSβ complex formation, although this difference was concentration. The fraction of expansions that are resistant to not statistically significant (SI Appendix,Fig.S4E and F). A third RGFP966 suggests that the deacetylation-mimic Msh3-5KR ob- β viates the need for HDAC3 and therefore bypasses the effects of mechanism would be if HDAC3 controls access of MutS to the triplet repeat DNA; however, previous work using chromatin the inhibitor. Thus, in the absence of RGFP966, HDAC3 appears β to directly deacetylate MutSβ to enhance its expansion activity. immunoprecipitation showed that MutS occupancy at the TNR was unaltered by siRNA knockdown of HDAC3 (39). HDAC3 Deacetylates Msh3 Peptides In Vitro. The hypothesis that An alternative explanation was suggested by the clustering of HDAC3 deacetylates MutSβ predicts that HDAC3 will be active the five key lysine residues in Msh3. Perhaps these lysine residues on acetylated peptides from Msh3. Five 15-residue peptides were serve as a NLS, and this NLS might be regulated by acetylation. If chemically synthesized with Msh3 protein sequence (Fig. 2A and so, acetylation-dependent partitioning of MutSβ between the nu- SI Appendix, Fig. S2). Each peptide contained a single acetylated cleus and cytoplasm would provide a mechanism to explain the lysine residue in the center. NMR of each peptide gave a peak connection between HDAC3 and triplet repeat expansions. near 1.83 ppm, characteristic of acetyl-lysine (Fig. 2A). This peak Alignment of a known NLS motif in the Msh6 subunit of MutSα disappeared after treatment with purified SMRT(389-480)/ (67, 68) with Msh3 and Msh3-5KR (Fig. 4A) showed that Msh3 HDAC3 (SI Appendix,Fig.S3B) (65), with a subsequent increase lysine residues 98 and 99 align with part of the NLS from Msh6, in free acetic acid near 1.78 ppm (Fig. 2A). Time course experi- and that Msh3 lysine 103 also aligns with a cognate lysine in Msh6. ments monitored by pseudo two-dimensional (2D) NMR showed The putative NLS in Msh3 overlaps with a strong functional NLS that each peptide was deacetylated by HDAC3 but with different between residues 98 and 107 of Msh3 that was recently described apparent rates (Fig. 2B). The K122ac peptide was deacetylated (69). We sought to build on this knowledge with experiments to most rapidly, followed by K99ac and K123ac at approximately test the possibility that the acetylation status of lysine residues in equal rates, then K98ac and K103ac. The difference in rate be- the predicted Msh3 NLS controls the subcellular localization of tween fastest and slowest rates was approximately fourfold. The Msh3 and thereby helps regulate triplet repeat expansions.

4of9 | www.pnas.org/cgi/doi/10.1073/pnas.2013223117 Williams et al. Downloaded by guest on September 27, 2021 BIOCHEMISTRY

− Fig. 2. Deacetylation assays of acetylated peptides derived from Msh3 in the presence of SMRT(389-480)/HDAC3 complex. (A) KAc and CH3COO peak integral quantification. Dynamics Center (Bruker) interface showing the alignment of peaks corresponding to the acetyl group of the acetylated lysines (substrate, ∼1.83 ppm) and the free acetic acid (product, ∼1.78 ppm) after incubation with SMRT(389-480)/HDAC3 at 20 °C for 60 min. (B) SMRT(389-480)/ HDAC3 complex preferentially deacetylates Msh3 K122 peptide over the rest of the substrates. The ratio of the integrals of the products (free acetic acid) over the integrals of the substrates (acetyl group of lysines) was plotted against reaction time.

Subcellular fractionation experiments tested whether MutSβ of the SV40 large T antigen NLS to EGFP resulted in nuclear lo- relocates, in whole or in part, from nucleus to cytoplasm when calization (Fig. 5A). Neither control fusion protein showed sub- HDAC3 is inhibited. In cells expressing wild-type Msh3, both stantial alteration in localization upon treatment with RGFP966. Msh3 and Msh2 were predominantly located in the nucleus Fusion of wild-type Msh391-130 to EGFP resulted in primarily nu- without any HDAC3 inhibitor (Fig. 4 B and D and SI Appendix, clear localization, but a significant amount of fluorescence was Fig. S1C). Only 2% and 15% of Msh3 and Msh2, respectively, shifted to the cytoplasm upon the addition of RGFP966. In con- fractionated with the cytoplasm. However, treatment of the cells trast, Msh3-5KR91-130-EGFP remained predominantly nuclear even with RGFP966 caused an additional 10% of both subunits to when the HDAC3 inhibitor was added (Fig. 5A). IF experiments shift to the cytoplasm (Fig. 4 B and D and SI Appendix, Fig. S1C). were also performed on cells with stably integrated, full-length fu- In contrast, cells expressing Msh3-5KR showed no detectable sions of wild-type Msh3-EGFP or Msh3-5KR-EGFP. Wild-type difference in cytoplasmic localization with or without RGFP966 Msh3-EGFP was predominantly nuclear with DMSO addition but (Fig. 4 C and E). These data offer an explanation for the ability some fluorescence was observed in the cytoplasm with RGFP966 of cells with Msh3-5KR to partially resist the effects of HDAC3 treatment (Fig. 5B). The extent of the relocalization appeared to be inhibition by partly blocking the relocalization of MutSβ to the in line with the subcellular localization experiments (Fig. 4). In cytoplasm. Immunofluorescence (IF) microscopy provided ad- contrast, the IF images with full-length Msh3-5KR-EGFP showed ditional support. We constructed fusions of Msh3 amino acid no apparent relocalization of nuclear fluorescence intensity with residues 91 to 130, the region containing the five lysine residues RGFP966 treatment (Fig. 5C). In total, the subcellular fractionation of interest, with EGFP. Transient transfection showed that an and IF images support a model where at least some MutSβ reloc- EGFP-only control was distributed throughout the cell, while fusion alizes from the nucleus to the cytoplasm with HDAC3 inhibition.

Williams et al. PNAS Latest Articles | 5of9 Downloaded by guest on September 27, 2021 This redistribution is mostly blocked by replacement of the five key lysine residues with arginines. Discussion This study reveals the biological consequence of HDAC3 activation of MutSβ in triplet repeat expansions and suggests a molecular mechanism connecting HDAC3 to DNA mutagenesis at disease- causing genes. Using the selective inhibitor RGFP966, we demon- strated that HDAC3 inhibition prevents triplet repeat expansion in human cells without hindering canonical mismatch repair activity. Importantly, expansions in cells with five lysine-to-arginine mutations in the Msh3 protein (Msh3-5KR) were largely insensitive to HDAC3 inhibition, consistent with the idea that some or all of the five lysine residues in Msh3 are key targets of HDAC3 deacetylation. The loss of some expansions in the presence of RGFP966 is presumably due to other sites of HDAC3 action, possibly additional residues in Msh3, Msh2, or in other proteins. We also show that purified SMRT(389-480)/HDAC3 complex deacetylates Msh3 peptide substrates containing the five lysines. One outcome of HDAC3 action is to partially control the nuclear localization of MutSβ. Immunofluorescence microscopy and subcellular localization approaches revealed that the deacetylation sites in Msh3 overlap a nuclear localization signal (69). Cells expressing wild-type Msh3 sequences showed nuclear localization that was reduced, although not eliminated, by addition of RGFP966. In contrast, expression of Msh3-5KR led to retention of protein nuclear localization even after treatment with the HDAC3 inhibitor. To- gether, our findings suggest an innovative regulatory mechanism for TNR expansion involving direct deacetylation of MutSβ by HDAC3. This work underscores HDAC3-selective inhibition as a potential therapy in Huntington’s disease and some related CAG•CTG expansion diseases. This study reports regulation of MutSβ by HDACs, specifically the class I enzyme HDAC3, assisted by the class IIa protein HDAC5 (39). In contrast, the Msh2-Msh6 complex MutSα is regulated by the class IIb enzyme HDAC6 through effects on Msh2 protein stability (70). A different class IIb enzyme, HDAC10, also interacts with Msh2, and MutSα-dependent MMR activity correlates positively with HDAC10 expression levels (71). The MutL homolog Mlh1 undergoes deacetylation by HDAC6 (72), which leads to blockage of assembly of MutSα- MutLα. In summary, there is precedent for class IIb HDAC- dependent regulation of MutSα and MutLα in canonical mismatch repair and response to DNA damage. Our work indicates that MutSβ is activated by class I HDAC3 and class IIa HDAC5 (39), leading to enhanced triplet repeat expansions. We note that the expansion outcome as a function of RGFP966 concentration appears to be dose dependent for Msh3 but less so for Msh3- 5KR (Fig. 1D). This could be a technical limitation in measuring expansion frequencies. Alternatively, it could be that deacetylation of other lysine sites outside of Msh3 is readily inhibited with RGFP966 at 1 μM concentration, thus blocking approximately one-half of expansions equally well for both Msh3 and Msh3- 5KR. In contrast, 10 μM RGFP966 is required to inhibit deacetylation of the five key sites in Msh3 to block the other approximate half of expansions. The Msh3-5KR substitutions would prevent the effect of higher inhibitor dose. The flattening of the response curve for Msh3-5KR could be explained by this possibility. ThefactthatMutSβ abundance and complex formation are not Fig. 3. DNA mismatch repair and DNA hairpin repair assays after RGFP966 altered by HDAC3 inhibition suggests an innovative regulatory treatment. (A) Western blots showing expression of MMR proteins and re- mechanism for TNR expansion involving, in part, HDAC3- ′ lated histone marks. (B) In vitro MMR assay using a 5 nicked heteroduplex mediated nuclear localization of MutSβ. This hypothesis is largely containing a G-T mismatch or a 2-nt insertion/deletion mispair. The % of α repair represents the average repair of three independent experiments. (C) consistent with studies on localization of MutS (68, 73, 74). Mouse and human Msh2 do not contain a putative NLS (73, 74), suggesting Diagram of DNA substrate containing a (CAG)5 or (CAG)15 hairpin structure α in the continuous strand. (D) Southern blot analysis showing nick-directed that MutS nuclear import is completely dependent on the Msh6 hairpin repair. (E) Quantification of hairpin repair. P < 0.04, n = 3. n.s., not subunit. Msh6 contains three NLSs that allow for MutSα import significant. See SI Appendix, Fig. S1C for statistical details. into the nucleus (68). The Msh6 NLSs are clustered in the N-terminal

6of9 | www.pnas.org/cgi/doi/10.1073/pnas.2013223117 Williams et al. Downloaded by guest on September 27, 2021 cytoplasm and the nucleus, allowing nuclear MutSβ access to triplet repeat DNA to drive expansions. This condition pre- dominates in the presence of the Msh3-5KR mutant that mimics the deacetylated state. In contrast, acetylated MutSβ can exit the nucleus but its reentry is inhibited. This reduces the level of nuclear MutSβ, limits its access to triplet repeat DNA, and helps suppress expansions. This condition is simulated when cells are treated with RGFP966. There is precedent for the idea that protein acetylation regulates nuclear localization. NLS acetylation has been identified as a molecular toggle for the IFN- inducible viral DNA sensor IFI16, inhibiting its nuclear import (75). Acetylation of a lysine in the NLS of c-Abl drives cytoplasmic accumulation (76). For other proteins, lysine acetylation is required for nuclear localization and retention, including the eukaryotic translation initiation factor 5A (77) and retinoblas- toma protein (78). The impact of acetylation on NLS function seems versatile, either promoting or preventing nuclear import. Another possibility for HDAC3/SMRT control of expansions is that the acetylation status of MutSβ affects its binding to triplet repeat hairpins or other alternative secondary structures that are thought to be intermediates in expansions (7–10). This explanation would help tie acetylation of MutSβ to structure-forming triplet repeats, which coincide with the subset of sequences known to BIOCHEMISTRY

Fig. 4. Subcellular fractionation of MutSβ following treatment with RGFP966. (A) Sequence alignment of a known NLS in Msh6 (67, 68) with Msh3 and Msh3-5KR. Msh3 residues 98, 99, and 103 are shown in boldface. (B) Representative Western blot of subcellular fractionation of MutSβ.(C) Representative Western blot of subcellular fractionation of MutSβ containing Msh3-5KR. (D) Quantification of subcellular fractionation experiments performed in Msh3-expressing cells. Graph depicts percentage of total Msh3 or Msh2 protein that is localized to the cytoplasm. *P = 0.0183, n = 3, SI Appendix, Fig. S1C.(E) Quantification of subcellular fractionation experiments performed in Msh3-5KR-expressing cells. Graph depicts percentage of total Msh3 or Msh2 protein that is localized to the cytoplasm, n = 3.

region of the protein and share some sequence homology with the N terminus of Msh3. It was this homology that guided our thinking with respect to potential NLS activity in Msh3 (Fig. 4A). Another group recently reported NLS activity in this same Msh3 motif (69). They showed that the strongest NLS in Msh3 is between residues 84 and 107, with particular requirement for the region 98 to 107. Our results are in agreement with this conclusion (SI Appendix,Fig.S6). We go on to show that NLS activity in MutSβ is regulated in part by HDAC3 deacetylation of five lysine residues between residues 98 and 123. One challenge in interpreting this information is that the N-terminal 220 amino acids of Msh3 appear to be largely unstructured, based on bioinformatics analysis (SI Appendix,Fig.S3C and D). As HDAC3 inhibition suppresses most expansions but its effect on MutSβ localization is only partial, we speculate there could be a subpopulation of MutSβ that is responsible for triplet repeat expansions that is also Fig. 5. Immunofluorescence microscopy using EGFP tags. In all panels, DAPI sensitive to cellular relocalization in response to changes in its + GFP indicates overlay of separate imaging for DAPI and EGFP. When acetylation status. present, RGFP966 dose was 10 μM. Scale bars are indicated in each image. (A) A model is presented to suggest how acetylation/deacetylation Transient transfections of EGFP fusion peptides. Top Left, EGFP alone; Top Right, SV40-NLS3x-EGFP positive control; Bottom Left, Msh391-130-EGFP; of MutSβ affects triplet repeat expansions (Fig. 6). Import of 91-130 β Bottom Right, Msh3-5KR -EGFP. FITC channel for EGFP signal. (B) Stable MutS into the nucleus via the lysine-rich Msh3 NLS allows transfectants expressing full-length Msh3-EGFP. White arrows indicate cy- acetylation by p300/CBP and deacetylation by SMRT/HDAC3. toplasmic accumulation of fluorescent protein. (C) Stable transfectants In this model, deacetylated MutSβ can shuttle freely between the expressing full-length Msh3-5KR-EGFP.

Williams et al. PNAS Latest Articles | 7of9 Downloaded by guest on September 27, 2021 Fig. 6. Model for SMRT/HDAC3 control of MutSβ subcellular localization. MutSβ is able to translocate into the nucleus due to the nuclear localization sequence in Msh3. Inside the nucleus, MutSβ is subject to acetylation by p300/CBP and deacetylation by SMRT/HDAC3. The deacetylated form of MutSβ can translocate in and out of the nucleus (Left). MutSβ containing Msh3-5KR simulates this situation. In contrast, the acetylated form of MutSβ can exit the nucleus (Right) but its reentry occurs with reduced efficiency (red vertical bar). Treatment with RGFP966 favors acetylated MutSβ, hinders its nuclear reentry, and thereby suppresses triplet repeat expansions.

expand and cause disease (79). Alternatively, protein-protein in- disease onset is determined, in part, by Msh3-dependent ex- teractions between MutSβ and other proteins involved in expan- pansions of CAG•CTG repeats (46). Additional support came sions might be subject to modulation by acetylation-deacetylation. from a study in HD mice that traced strain-dependent differ- Developing suitable assays for these activities presents an interesting ences in CAG somatic repeat instability to polymorphisms in challenge going forward. MSH3 primarily affecting Msh3 protein stability (35). Direct While triplet repeat expansions were very sensitive to activation of MutSβ by HDAC3 also provides a mechanistic ex- RGFP966 inhibition of HDAC3, canonical mismatch repair and planation for how RGFP966 treatment of HD mice suppressed CAG loop repair activities were slightly stimulated by the in- striatal instability in the HTT CAG repeat (53). It is a reasonable hibitor. The most likely explanation is that MutSα, the pre- inference that RGFP966 inhibition of HDAC3 in vivo prevented dominant complex in HeLa cells (80), is not affected by HDAC3 β inhibition and thereby masks any effects on the lower-abundance deacetylation of Msh3 and thereby blocked MutS in driving MutSβ in these repair assays. This interpretation is consistent repeat expansion. Therefore, a direct mechanistic connection is with the known dependence of G-T repair on MutSα and the feasible between HDAC3 inhibition and slowing somatic repeat overlap in repair activity between MutSα and MutSβ for 2-nt expansion to alter the course of triplet repeat expansion disease. looped substrates (80, 81). CAG loop repair does not require MutSβ (63, 82). Nonetheless, the retention of canonical repair ac- Data Availability. All study data are included in the article and tivity after HDAC3 inhibition suggests there may be no genetic price SI Appendix. to pay in therapeutic approaches to expansions using RGFP966. Our findings are also relevant to recent GWAS findings that ACKNOWLEDGMENTS. This work was supported by Biotechnology and Bio- genetic variation in MSH3 influences somatic expansions and logical Sciences Research Council (BBSRC)-Science Foundation Ireland Joint – MSH3 Funding of Research Award 16/BBSRC/3395 (to J.W.R.S. and R.S.L.), by the Irish disease severity in HD and DM1 (43 46). Increased ex- Research Council Government of Ireland Postdoctoral Fellowship (to G.M.W.), pression is associated with disease progression in both diseases and by the Cancer Prevention and Research Institute of Texas (CPRIT) Award (44, 46), helping support the conclusion that the timing of RR160101 (to G.-M.L.). G.-M.L. is a CPRIT Scholar in Cancer Research.

1. The Huntington’s Disease Collaborative Research Group, A novel gene containing a 14. M. Zatz et al., Analysis of the CTG repeat in skeletal muscle of young and adult trinucleotide repeat that is expanded and unstable on Huntington’s disease chro- myotonic dystrophy patients: When does the expansion occur? Hum. Mol. Genet. 4, mosomes. Cell 72, 971–983 (1993). 401–406 (1995). 2. J. D. Brook et al., Molecular basis of myotonic dystrophy: Expansion of a trinucleotide 15. L. Martorell et al., Germline mutational dynamics in myotonic dystrophy type 1 males: (CTG) repeat at the 3′ end of a transcript encoding a protein kinase family member. Allele length and age effects. Neurology 62, 269–274 (2004). Cell 68, 799–808 (1992). 16. C. Savouret et al., MSH2-dependent germinal CTG repeat expansions are produced 3. Y.-H. Fu et al., An unstable triplet repeat in a gene related to myotonic muscular continuously in spermatogonia from DM1 transgenic mice. Mol. Cell. Biol. 24, dystrophy. Science 255, 1256–1258 (1992). 629–637 (2004). 4. M. Mahadevan et al., Myotonic dystrophy mutation: An unstable CTG repeat in the 3′ 17. H. Telenius et al., Somatic and gonadal mosaicism of the Huntington disease gene untranslated region of the gene. Science 255, 1253–1255 (1992). CAG repeat in brain and sperm. Nat. Genet. 6, 409–414 (1994). 5. E. J. Kremer et al., Mapping of DNA instability at the fragile X to a trinucleotide re- 18. C. A. Thornton, K. Johnson, R. T. Moxley 3rd, Myotonic dystrophy patients have larger peat sequence p(CCG)n. Science 252, 1711–1714 (1991). CTG expansions in skeletal muscle than in leukocytes. Ann. Neurol. 35, 104–107 6. A. J. Verkerk et al., Identification of a gene (FMR-1) containing a CGG repeat coin- (1994). cident with a breakpoint cluster region exhibiting length variation in fragile X syn- 19. M. T. Fortune, C. Vassilopoulos, M. I. Coolbaugh, M. J. Siciliano, D. G. Monckton, drome. Cell 65, 905–914 (1991). Dramatic, expansion-biased, age-dependent, tissue-specific somatic mosaicism in a 7. K. Usdin, N. C. House, C. H. Freudenreich, Repeat instability during DNA repair: In- transgenic mouse model of triplet repeat instability. Hum. Mol. Genet. 9, 439–445 sights from model systems. Crit. Rev. Biochem. Mol. Biol. 50, 142–167 (2015). (2000). 8. R. R. Iyer, A. Pluciennik, M. Napierala, R. D. Wells, DNA triplet repeat expansion and 20. I. V. Kovtun, C. T. McMurray, Trinucleotide expansion in haploid germ cells by gap mismatch repair. Annu. Rev. Biochem. 84, 199–226 (2015). repair. Nat. Genet. 27, 407–411 (2001). 9. M. H. M. Schmidt, C. E. Pearson, Disease-associated repeat instability and mismatch 21. C. Savouret et al., CTG repeat instability and size variation timing in DNA repair- repair. DNA Repair (Amst.) 38, 117–126 (2016). deficient mice. EMBO J. 22, 2264–2273 (2003). 10. A. N. Khristich, S. M. Mirkin, On the wrong DNA track: Molecular mechanisms of 22. R. Gonitel et al., DNA instability in postmitotic neurons. Proc. Natl. Acad. Sci. U.S.A. repeat-mediated genome instability. J. Biol. Chem. 295, 4134–4170 (2020). 105, 3467–3472 (2008). 11. E. P. Leeflang et al., Analysis of germline mutation spectra at the Huntington’s 23. M. Gomes-Pereira, M. T. Fortune, D. G. Monckton, Mouse tissue culture models of disease locus supports a mitotic mutation mechanism. Hum. Mol. Genet. 8,173–183 unstable triplet repeats: In vitro selection for larger alleles, mutational expansion bias (1999). and tissue specificity, but no association with cell division rates. Hum. Mol. Genet. 10, 12. L. Zhang, E. P. Leeflang, J. Yu, N. Arnheim, Studying human mutations by sperm 845–854 (2001). typing: Instability of CAG trinucleotide repeats in the human androgen receptor 24. M. Gomes-Pereira et al., Disease-associated CAG·CTG triplet repeats expand rapidly in gene. Nat. Genet. 7, 531–535 (1994). non-dividing mouse cells, but cell cycle arrest is insufficient to drive expansion. Nucleic 13. E. P. Leeflang et al., Single sperm analysis of the trinucleotide repeats in the Hun- Acids Res. 42, 7047–7056 (2014). tington’s disease gene: Quantification of the mutation frequency spectrum. Hum. 25. M.-Y. Chung et al., Evidence for a mechanism predisposing to intergenerational CAG Mol. Genet. 4, 1519–1526 (1995). repeat instability in spinocerebellar ataxia type I. Nat. Genet. 5, 254–258 (1993).

8of9 | www.pnas.org/cgi/doi/10.1073/pnas.2013223117 Williams et al. Downloaded by guest on September 27, 2021 26. G. M. Goellner et al., Different mechanisms underlie DNA instability in Huntington 54. D. A. Claassen, R. S. Lahue, Expansions of CAG.CTG repeats in immortalized human disease and colorectal cancer. Am. J. Hum. Genet. 60, 879–890 (1997). astrocytes. Hum. Mol. Genet. 16, 3088–3096 (2007). 27. K. Manley, T. L. Shirley, L. Flaherty, A. Messer, Msh2 deficiency prevents in vivo so- 55. G. M. Williams, R. S. Lahue, Assessing triplet repeat expansions in human SVG-A cell matic instability of the CAG repeat in Huntington disease transgenic mice. Nat. Genet. culture. Methods Mol. Biol. 2056, 151–172 (2020). 23, 471–473 (1999). 56. M. Malvaez et al., HDAC3-selective inhibitor enhances extinction of cocaine-seeking 28. W. J. A. A. van den Broek et al., Somatic expansion behaviour of the (CTG)n repeat in behavior in a persistent manner. Proc. Natl. Acad. Sci. U.S.A. 110, 2647–2652 (2013). myotonic dystrophy knock-in mice is differentially affected by Msh3 and 57. H. Jia et al., The effects of pharmacological inhibition of histone deacetylase 3 Msh6 mismatch-repair proteins. Hum. Mol. Genet. 11, 191–198 (2002). (HDAC3) in Huntington’s disease mice. PLoS One 11, e0152498 (2016). 29. V. C. Wheeler et al., Mismatch repair gene Msh2 modifies the timing of early disease 58. J. D. Dignam et al., Accurate transcription initiation by RNA polymerase II in a soluble in Hdh(Q111) striatum. Hum. Mol. Genet. 12, 273–281 (2003). extract from isolated mammalian nuclei. Nucleic Acids Res. 11, 1475–1489 (1983). 30. B. A. L. Owen et al., (CAG)(n)-hairpin DNA binds to Msh2-Msh3 and changes prop- 59. J. Holmes Jr., S. Clark, P. Modrich, Strand-specific mismatch correction in nuclear ex- erties of mismatch recognition. Nat. Struct. Mol. Biol. 12, 663–670 (2005). tracts of human and Drosophila melanogaster cell lines. Proc. Natl. Acad. Sci. U.S.A. 31. L. Foiry et al., Msh3 is a limiting factor in the formation of intergenerational CTG 87, 5837–5841 (1990). expansions in DM1 transgenic mice. Hum. Genet. 119, 520–526 (2006). 60. S. Guo et al., Differential requirement for proliferating cell nuclear antigen in 5′ and 32. E. Dragileva et al., Intergenerational and striatal CAG repeat instability in Hunting- 3′ nick-directed excision in human mismatch repair. J. Biol. Chem. 279, 16912–16917 ’ ton s disease knock-in mice involve different DNA repair genes. Neurobiol. Dis. 33, (2004). – + 37 47 (2009). 61. R. Parsons et al., Hypermutability and mismatch repair deficiency in RER tumor cells. 33. R. A. Lokanga, X. N. Zhao, K. Usdin, The mismatch repair protein MSH2 is rate limiting Cell 75, 1227–1236 (1993). for repeat expansion in a fragile X premutation mouse model. Hum. Mutat. 35, 62. J. Ortega et al., Phosphorylation of PCNA by EGFR inhibits mismatch repair and – 129 136 (2014). promotes misincorporation during DNA synthesis. Proc. Natl. Acad. Sci. U.S.A. 112, β 34. X.-N. Zhao et al., Muts generates both expansions and contractions in a mouse 5667–5672 (2015). – model of the Fragile X-associated disorders. Hum. Mol. Genet. 24, 7087 7096 (2015). 63. T. Zhang, J. Huang, L. Gu, G. M. Li, In vitro repair of DNA hairpins containing various 35. S. Tomé et al., MSH3 polymorphisms and protein levels affect CAG repeat instability numbers of CAG/CTG trinucleotide repeats. DNA Repair (Amst.) 11, 201–209 (2012). ’ in Huntington s disease mice. PLoS Genet. 9, e1003280 (2013). 64. P. V. Hornbeck et al., PhosphoSitePlus, 2014: Mutations, PTMs and recalibrations. β β 36. X. N. Zhao et al., A MutS -dependent contribution of MutS to repeat expansions in Nucleic Acids Res. 43, D512–D520 (2015). fragile X premutation mice? PLoS Genet. 12, e1006190 (2016). 65. P. J. Watson, L. Fairall, G. M. Santos, J. W. Schwabe, Structure of HDAC3 bound to co- 37. A. Kantartzis et al., Msh2-Msh3 interferes with Okazaki fragment processing to repressor and inositol tetraphosphate. Nature 481, 335–340 (2012). – promote trinucleotide repeat expansions. Cell Rep. 2, 216 222 (2012). 66. Z. A. Gurard-Levin, K. A. Kilian, J. Kim, K. Bähr, M. Mrksich, Peptide arrays identify 38. G. M. Williams, J. A. Surtees, MSH3 promotes dynamic behavior of trinucleotide re- isoform-selective substrates for profiling endogenous lysine deacetylase activity. ACS peat tracts in vivo. Genetics 200, 737–754 (2015). Chem. Biol. 5, 863–873 (2010). 39. A.-M. M. Gannon, A. Frizzell, E. Healy, R. S. Lahue, MutSβ and histone deacetylase 67. M. Christmann, B. Kaina, Nuclear translocation of mismatch repair proteins MSH2 and complexes promote expansions of trinucleotide repeats in human cells. Nucleic Acids MSH6 as a response of cells to alkylating agents. J. Biol. Chem. 275, 36256–36262 Res. 40, 10324–10333 (2012). (2000). 40. J. Du, E. Campau, E. Soragni, C. Jespersen, J. M. Gottesfeld, Length-dependent CTG· 68. N. R. Gassman et al., Cooperative nuclear localization sequences lend a novel role to CAG triplet-repeat expansion in myotonic dystrophy patient-derived induced plu- the N-terminal region of MSH6. PLoS One 6, e17907 (2011). ripotent stem cells. Hum. Mol. Genet. 22, 5276–5287 (2013). BIOCHEMISTRY 69. S. S. Tseng-Rogenski et al., The human DNA mismatch repair protein MSH3 contains 41. R. Nakatani, M. Nakamori, H. Fujimura, H. Mochizuki, M. P. Takahashi, Large ex- nuclear localization and export signals that enable nuclear-cytosolic shuttling in re- pansion of CTG•CAG repeats is exacerbated by MutSβ in human cells. Sci. Rep. 5, sponse to inflammation. Mol. Cell. Biol. 40, e00029-20 (2020). 11020 (2015). 70. M. Zhang et al., HDAC6 deacetylates and ubiquitinates MSH2 to maintain proper 42. N. Keogh, K. Y. Chan, G. M. Li, R. S. Lahue, MutSβ abundance and Msh3 ATP hydrolysis levels of MutSα. Mol. Cell 55,31–46 (2014). activity are important drivers of CTG•CAG repeat expansions. Nucleic Acids Res. 45, 71. R. Radhakrishnan et al., Histone deacetylase 10 regulates DNA mismatch repair and 10068–10078 (2017). may involve the deacetylation of MutS homolog 2. J. Biol. Chem. 290, 22795–22804 43. D. J. H. Moss et al.; TRACK-HD investigators; REGISTRY investigators, Identification of (2015). genetic variants associated with Huntington’s disease progression: A genome-wide 72. M. Zhang et al., HDAC6 regulates DNA damage response via deacetylating MLH1. association study. Lancet Neurol. 16, 701–711 (2017). J. Biol. Chem. 294, 5813–5826 (2019). 44. M. Flower et al., MSH3 modifies somatic instability and disease severity in Hunting- 73. N. O. Knudsen et al., Nuclear localization of human DNA mismatch repair protein ton’s and myotonic dystrophy type 1. Brain 142, 1876–1886 (2020). – 45. M. Ciosi et al.; TRACK-HD team; Enroll-HD team, A genetic association study of exonuclease 1 (hEXO1). Nucleic Acids Res. 35, 2609 2619 (2007). glutamine-encoding DNA sequence structures, somatic CAG expansion, and DNA 74. N. O. Knudsen, S. D. Andersen, A. Lützen, F. C. Nielsen, L. J. Rasmussen, Nuclear repair gene variants, with Huntington disease clinical outcomes. EBioMedicine 48, translocation contributes to regulation of DNA excision repair activities. DNA Repair – 568–580 (2019). (Amst.) 8, 682 689 (2009). 46. Genetic Modifiers of Huntington’s Disease (GeM-HD) Consortium, CAG repeat not 75. T. Li, B. A. Diner, J. Chen, I. M. Cristea, Acetylation modulates cellular distribution and polyglutamine length determines timing of Huntington’s disease onset. Cell 178, DNA sensing ability of interferon-inducible protein IFI16. Proc. Natl. Acad. Sci. U.S.A. – 887–900.e14 (2019). 109, 10558 10563 (2012). 47. Genetic Modifiers of Huntington’s Disease (GeM-HD) Consortium, Identification of 76. M. G. di Bari et al., c-Abl acetylation by histone acetyltransferases regulates its – genetic factors that modify clinical onset of Huntington’s disease. Cell 162, 516–526 nuclear-cytoplasmic localization. EMBO Rep. 7, 727 733 (2006). (2015). 77. M. Ishfaq et al., Acetylation regulates subcellular localization of eukaryotic transla- – 48. L. Jones, H. Houlden, S. J. Tabrizi, DNA repair in the trinucleotide repeat disorders. tion initiation factor 5A (eIF5A). FEBS Lett. 586, 3236 3241 (2012). Lancet Neurol. 16,88–96 (2017). 78. A. Pickard, P. P. Wong, D. J. McCance, Acetylation of Rb by PCAF is required for nu- 49. F. Morales et al., A polymorphism in the MSH3 mismatch repair gene is associated clear localization and keratinocyte differentiation. J. Cell Sci. 123, 3718–3726 (2010). with the levels of somatic instability of the expanded CTG repeat in the blood DNA of 79. A. M. Gacy, G. Goellner, N. Juranic, S. Macura, C. T. McMurray, Trinucleotide repeats myotonic dystrophy type 1 patients. DNA Repair (Amst.) 40,57–66 (2016). that expand in human disease form hairpin structures in vitro. Cell 81, 533–540 (1995). 50. K. Debacker et al., Histone deacetylase complexes promote trinucleotide repeat ex- 80. J. T. Drummond, J. Genschel, E. Wolf, P. Modrich, DHFR/MSH3 amplification in pansions. PLoS Biol. 10, e1001257 (2012). methotrexate-resistant cells alters the hMutSalpha/hMutSbeta ratio and reduces the 51. J. Oberoi et al., Structural basis for the assembly of the SMRT/NCoR core transcrip- efficiency of base-base mismatch repair. Proc. Natl. Acad. Sci. U.S.A. 94, 10144–10149 tional repression machinery. Nat. Struct. Mol. Biol. 18, 177–184 (2011). (1997). 52. G. M. Hudson, P. J. Watson, L. Fairall, A. G. Jamieson, J. W. Schwabe, Insights into the 81. J. Genschel, S. J. Littman, J. T. Drummond, P. Modrich, Isolation of MutSbeta from recruitment of Class IIa histone deacetylases (HDACs) to the SMRT/NCoR transcrip- human cells and comparison of the mismatch repair specificities of MutSbeta and tional repression complex. J. Biol. Chem. 290, 18237–18244 (2015). MutSalpha. J. Biol. Chem. 273, 19895–19901 (1998). 53. N. Suelves et al., An HDAC3-selective inhibitor delivers concurrent benefits in Hun- 82. G. B. Panigrahi, R. Lau, S. E. Montgomery, M. R. Leonard, C. E. Pearson, Slipped (CTG) tington’s disease mice by preventing cognitive decline and suppressing somatic CAG *(CAG) repeats can be correctly repaired, escape repair or undergo error-prone re- repeat expansions. Sci. Rep. 7, 6082 (2017). pair. Nat. Struct. Mol. Biol. 12, 654–662 (2005).

Williams et al. PNAS Latest Articles | 9of9 Downloaded by guest on September 27, 2021